2.1 Data Dissemination Mechanisms

The exchange of data between vehicles is an important part of applications deployed in VANETs, and its management has a real impact on their performance and functionality. When deployed on mobile and heterogeneous environments, in circumstances of high or low intensity of network traffic, the way in which data is disseminated faces different challenges. Here we present a body of research work that has been devoted to solve the problem of the exchange and propagation of relevant safety information in a vehicular network.

The concept of data dissemination often refers to data-casting, which can be seen as a limited and controlled way in which data flows through the network [¹⁰]. Unlike routing algorithms, which seek to discover a route to deliver information between two nodes, data dissemination strategies are employed to deliver a message to all nodes or a particular group, located in a specific geographical area. Therefore, data dissemination takes into account the network load and at the same time it must ensure to deliver data to nodes that need it, with reduced latency, to facilitate the decision-making process in a timely manner.

In the past, there have been schemes and protocols for dissemination that rely on the infrastructure or on an ad-hoc network [¹¹, ¹²]. The review in this section focuses only on works that fall on the ad-hoc category for dissemination, which resembles a VANET. A widely used mechanism for dissemination is the data delivery over multi-hop communications, where data is disseminated using intermediate nodes as relays [⁷]. However, a prevalent problem in the ad hoc networks is the so-called “broadcast storm” [⁸], which occurs when many vehicles send packets at the same time. In the presence of a large number of sources of traffic, packet collisions occur, decreasing quality in communication and generating data losses. In a VANET, such a situation causes that critical events, like near car accidents, are not disseminated on time. Therefore, the poor performance of the communications network represents a severe problem for the safety of users in the vehicular network, especially if we consider that a prompt reaction to risky events depends on the timely delivery of safety information. The problem of a congested communications network is usually related to a highly dense vehicular network traffic, which is a recurrent scenario in urban environments. The work in [¹³] is designed to avoid “broadcast storms” in vehicular scenarios.

In multi-hop dissemination, there are several proposals, some aimed at making the forwarding opportunistically [¹⁴]; other approaches employ genetic algorithms to choose the best position to perform multi-hop transmission, looking for the optimum configuration parameters with other methodologies such as fuzzy sets based on rules [¹⁵]. There are also protocols that consider scenarios of distant vehicles where the stability of network connectivity is very low. This area of study is known as Delay Tolerant Networks (DTN), which exploit opportunities for encounters in partitioned networks [¹⁶, ¹⁷]. In addition, the frequent partitions in VANETs are not only because nodes may be far away, but also due to obstacles that impede line of sight and produce adverse effects such as shadowing [¹⁸, ¹⁹], generating temporary disconnections among nodes.

Furthermore, some protocols establish a scheme to deal with scenarios with different vehicular traffic conditions [²⁰, ²¹], as well as schemes dedicated to events of interest in urban environments [²²]. Other works focus on providing reliable accident warnings on the route [²³] or they are designed for specific roads, such as intersections [⁵]. An example of reliable dissemination designed only to work on the first hop is described in [²⁴].

To address the problem of having different scenarios in the topology of VANETs, various forms of one-hop dissemination have been proposed. Such proposals for “adaptive beaconing” [²⁵, ²⁶] use the operating conditions to adapt the beacon rate, transmission power, and the transmission channel. An innovative proposal focused on adapting the same parameters using the variation of acceleration to decide how the “beaconing” mechanism is executed is described in reference [²⁷]. According to the parameters that the adaptive protocols modify, these can be classified into the following general categories [²⁸]:

Regarding the evaluation of the dissemination mechanisms, one can take into account different parameters to compare the effectiveness/efficiency among the protocols. Some of the most popular metrics used are the normalized routing load, packet losses, transmission rate, and latency. In [²⁹], a comparative study of dissemination protocols using “broadcasting” is presented. Although not very relevant for VANET scenarios, the energy efficiency has also been studied for some dissemination mechanisms to be able to implement them on intelligent transport systems that integrate other types of energy-efficient vehicles [³⁰].

2.2 Context-aware Systems

Context-aware systems are characterized by being aware or being sensitive of the situation (or context) in which they operate, either in physical, virtual or user’s environment [³¹].

The context is understood as “any information that can be used to characterize the situation of an entity. An entity is a person, place or object that is considered relevant to the interaction between a user and an application, including the user and the applications themselves” [³²]. These systems are able to adapt, exploiting the knowledge acquired through sensors, other nodes or through the infrastructure available on the route.

One of the first works related to context-aware systems was presented in [³³]. In this article, the authors developed an active tag location system, which based on infrared technology can determine the current location of a user, this was used to forward telephone calls to a telephone near the user.

In vehicular environments, context-aware systems focus mainly on the area of service recognition and location on the route [³⁴, ³⁵]. In [³²], the authors describe various approaches that have used context-aware systems for the detection of specific events, for example in security, entertainment, and traffic management applications. To the best of our knowledge, there is only one work that has addressed the problem of disseminating information of accidents based on warning messages considering the information of the vehicular context [³⁶]; however, the scheme does not consider the traffic of the communications network in the context definition.

Context-aware routing in VANETs is another area of research related to our work. In [³⁷], the authors present an approach to provide Internet access to support interactive and entertainment applications in VANETs. In [³⁸], the authors employ the source and destination location of nodes as the context information; then, they elaborate a routing mechanism that seeks to reduce packet losses due to saturation in the cache of each vehicle while avoiding network overload. In [³⁹], the authors define the vehicular context based on parameters like speed of each node and inter-vehicular distance. This context information is evaluated and taken into account while selecting the appropriate transmission speed considering the conditions of the link layer. The approach produces important VANET performance enhancements in terms of data transmission rates between vehicles.

In [⁴⁰], the authors present a classification of different broadcasting mechanisms and their effect in the performance of different VANET applications. The main contribution consists in the differentiation of broadcasting schemes based on client applications. Authors in [⁴¹] describe a context-aware system that identifies potential accidents and pro-actively enables safety mechanisms before the accident occurs. The information acquired from the sensors in the vehicles (i.e., data prior to the accident) is analyzed to predict a certain collision and its severity; in this way, the system launches a timely warning. In [⁴²], the authors propose a context-aware system to improve the way in which information is acquired to avoid redundancy in the information collected by a group of nodes, whereas the authors in [⁴³] propose a MAC layer context-aware protocol based on Bayesian networks with the aim at enhancing the performance of the VANET in highly congested scenarios. The approach presented in [⁴³] produces a substantial improvement compared to the CSMA/CA mechanism adopted by the IEEE 802.11-OCB standard, whose performance is seriously affected in highly congested environments.

3.1 Dissemination mechanisms chosen for the evaluation

• Slotted 1-persistence [¹³]

Is a dissemination mechanism that gives the priority of re-transmission to the farthest node of the sending node. For that, the protocol uses a waiting time before re-transmitting, which varies according to the distance between nodes. When a node receives a packet, the protocol verifies the packet ID; if the packet is received for the first time and if the node has not received any duplicates before, the protocol assigns a time slot. Then, the node performs a re-broadcast with a probability of 1 within the time slot Ts _ij assigned. Otherwise, the packet is discarded. The time slot can be obtained as indicated by Equation (1):

where τ is the estimated delay of a hop, which refers to the sum of the propagation delay plus the delay of accessing the shared medium. S _ij is the assigned slot number and is expressed in Equation (2) as follows:

where D _ij is the relative distance between the nodes i and j, R is the transmission range, and N _S is the default slot number. N _S must be chosen as a function of the traffic density. The authors in [²⁷] suggest a value of N _S = 5 in rush-hours or under traffic congestion scenarios. The value N _S = 3 is recommended for other cases. This protocol has been chosen for the study because it solves the problem of broadcast storms and because it needs only the information about the position of the sending and receiving nodes to operate. It is also one of the most widespread used protocols in the literature for its simplicity and because it was one of the first protocols proposed for dissemination.

• The Traffic Adaptive data Dissemination (TrAD) protocol [²¹]

This protocol employs adaptive beaconing, so that, based on the scenario in which mobiles nodes operate (urban or highway), the mechanism chooses between two ways of broadcasting that intend to improve the performance of the network. In addition, this mechanism assigns certain roles to the vehicles, depending on their position with respect to a place of interest, specifically when using store-carry-and-forward.

To avoid broadcast storms, TraD assigns different delays to re-transmissions of WAVE short messages (WSM) in the following way [²¹]: first, the source node establishes a ranking based on the utility function described with Equation (3) expressed as follows:

which involves the distance between sender and receiver D, the number of neighbors N of a given node, and the channel busy ratio CBR. D is calculated with Equation (4) expressed as follows:

Where Dist _sn is the distance between the sender and its neighbor and max.RadioRange is the maximum communication range over the wireless link.

N is calculated with Equation (5) expressed as follows:

where #Neighbor is the number of neighbors in the vicinity and max.#Neighbor is a pre-set maximum number of neighbors.

The ω _CBR is defined as a piece-wise linear function in terms of the CBR. It is defined in Equation (6) as follows:

The sending node collects the aforementioned information via beacons, i.e., each vehicle appends it own information before sending a beacon message. With this, a sender node sends the ranking to its neighbors. Then, upon reception of a safety message, the receiving node checks the ranking and assigns the delay for re-transmission Delay _TX . According to [²¹], Delay _TX is calculated in Equation (7) as:

where R corresponds to the position in the generated ranking. This protocol is chosen for the study since it takes into account the performance of the communications network as well as the state of the vehicular traffic.

3.2 Methodology and Characterization of scenarios

To obtain the baseline results, we employed the intersection illustrated in Figure 1, which is deployed on an intersection of two avenues with high vehicular traffic in the city of Santiago, Chile. The intersection has 3 and 4 lanes per road. The map is set in the SUMO simulator and we introduced vehicles according to 2 different densities: low 15.6 [veh/km/ln] and high 73.4 [veh/km/ln]. The densities seek to replicate representative values within fuzzy sets reported in [⁴⁴], where the authors classify different vehicular traffic density degrees, according to real measurements carried out in [⁴⁵]. With the objective to observe the performance at critical levels, we chose the extreme densities (low and high), which are consistent with the density values employed for the protocols under evaluation in their original proposal [¹³].

The scenario is evaluated with different configurations of the following network parameters:

• Beaconing frequency varying between 2 Hz and 10 Hz. These messages carry mandatory information for the operation of safety applications, such as position, velocity, and acceleration of the vehicle.

Figure 1 Views of scenario 

• Wave Service Advertisement messages (WSA) with frequency of 1 Hz. These messages are included to simulate background traffic on the network corresponding to various on-route services (i.e., traffic from other safety applications). They are additional load to the beaconing that operates permanently.

Table 1 Network loads definition 

The labeling of each scenario according to the existent load in the communications network is listed in Table 1. The loads were chosen according to the common frequency values employed in active road safety and cooperative navigation applications reported in [⁴⁶]. To understand the response of the mechanisms under study when there is an increase in network traffic load, an additional load coming from the WSA messages is added to the channel. Such a load is only included for the 10 Hz configuration to recreate extreme conditions of congestion over the channel.

During the simulation, a notification message for dissemination is generated only once by the source vehicle, that is, only one packet of type Wave Short Message is generated to observe its behavior at the first instant of dissemination. The packet is disseminated at time instant t = 5 s when an accident over the intersection triggers the safety notification message. The message is forwarded by the neighboring nodes across the entire region of interest (ROI). Because of this, when the PDR is calculated, this percentage is also a representation of the number of nodes in the ROI that received the message successfully.

We executed 20 runs for each configuration scenario, for a total of 320 simulation runs. Table 2 shows the details of the parameters used in the bidirectional simulator VEINS [⁴⁷] employed for this evaluation.

3.3 Discussion of results

To begin our discussion, we examine the Slotted 1-persistence’s performance; it should be noted that to the best of our knowledge, the effects of other values of N _s (see Equation (2)] have not been evaluated in the literature for particularly dense scenarios. The values N _s = 3 and N _s = 5 are the ones suggested by the authors of this mechanism for low- and high-density scenarios, respectively [¹³]. In this section, we present an evaluation of the protocol’s performance considering two additional values of N _s ; in low density scenarios we employ N _s = 2 and N _s = 3; in high density scenarios we employ N _s = 5 and N _s = 7.

In the case of a low density, we can observe in the results in Figure 2 the different behaviors of the protocol for different network loads. In particular, Figures 2a and 2b show that when the network load is high (i.e., 10 Hz of beaconing frequency and other WSA background messages) the PDR is reduced by almost a 40% and the end-to-end delay increases by almost a 200%. Although it seems the dissemination speed, shown in Figure 2c, increases in the same scenario, it has to be analyzed together with the PDR, i.e., most likely there is a high dissemination speed but for a reduced number of successfully delivered packets. Also, one could suggest that using N _s = 2 is better for any given context with a low vehicular density, independent of the network load. Although the average for PDR, end to end delay, and dissemination speed are very similar in both N _s = 2 and N _s = 3, it is shown that data dispersion is better in N _s = 2.

Figure 2 Slotted 1-persistence evaluation: different network loads and low vehicular density λ = 15.6 [veh/km/ln] 

Figure 3 Slotted 1-persistence evaluation: different network loads and high vehicular density λ = 73.4 [veh/km/ln] 

In the case of high density, one can observe in Figure 3 that when employing N _s = 7 instead of N _s = 5 it may result in a better protocol’s performance when the dense network is combined with network load configurations 1 and 2. In such scenarios, for N _s = 7, the PDR shown in Figure 3a increases around a 10% in average, with a reduced end to end delay and an higher dissemination speed (see Figures 3b and 3c). However, in the case of network load configuration 3, the dissemination speed decreases near a 60% in average, compared to N _s = 5, in which case it would not be the recommended value if applications have a strict requirement about the speed of dissemination.

Figure 4 corresponds to the evaluation of the TrAD protocol. We tested the protocol for the two vehicular densities and the different network loads. Note that no parameter values of the protocol are adjusted in the evaluation since the protocol employs an adaptive beaconing technique integrated into the protocol. The results show the high capacity of the protocol to adapt to different scenarios of operation, which delivers good results even when the scenario becomes more challenging (high density/high network load). However, this protocol involves additional complexity since it requires more computational capacity in each node and additional overhead appended to the beacon messages.

Figure 4 TrAD evaluation: different network loads and low/high vehicular density 

Table 2 Simulation parameters